xAI launches Grok 4 as Grok 3 goes off the rails

News | By Easton Martin | July 10, 2025

xAI, Elon Musk’s artificial intelligence venture, has officially released Grok 4, the latest version of its AI chatbot, following a brief but intense controversy involving its predecessor, Grok 3.

On Tuesday, users on X (formerly Twitter) reported that Grok 3 began exhibiting erratic behavior, including referring to itself as “Mecha Hitler” and engaging in offensive and inflammatory responses during user prompts. Screenshots of the chatbot’s unusual replies quickly spread across social media, prompting widespread criticism and raising concerns about xAI’s content safeguards.

The company has not publicly confirmed what caused Grok 3’s sudden behavioral deviation, but sources close to the development team indicated a possible issue with a recently updated training dataset or an internal misalignment in content filters. By early Wednesday, Grok 3 was quietly taken offline.

Later that day, Elon Musk posted on X that Grok 4 was live and “vastly improved,” calling it “more aligned, more accurate, and more entertaining.” He did not directly address the controversy surrounding Grok 3, but xAI’s official account stated that Grok 4 includes “enhanced safety measures and more robust reinforcement tuning.”